SPECT VOLUME QUANTITATION - INFLUENCE OF SPATIAL-RESOLUTION, SOURCE SIZE AND SHAPE, AND VOXEL SIZE

A number of factors influence the accuracy of estimation of source volume with single-photon emission computed tomography (SPECT) imaging. This study investigated the role of a number of factors including system spatial resolution (which includes the influence of low-pass filters applied to suppress noise), source size and shape, and voxel size in determining volume. A rectangular parallelepiped (bar), a right cylinder, and a sphere were mathematically modeled as being imaged with a SPECT system by calculating the three-dimensional (3-D) convolution of them with symmetric Gaussian functions of 20 different full widths at half maximums (FWHM's). The resulting activity profiles were analyzed to determine the location of the edges as a function of the source size relative to the FWHM of the system. The edge definition criteria studied were (1) the location of the 50% count threshold and (2) the maximum in the local gradient. In addition, the threshold which yielded the correct edge location was also determined. A nonstationary computer simulation of SPECT imaging, based on the serial model of the system transfer function, was used to test the predictions of the mathematical model and investigate the influence of (1) voxel size and sampling with a discrete array of voxels; (2) attenuation; (3) scatter; (4) variable spatial resolution; (5) low-pass filtering; and (6) noise. The mathematical model predicted that both the 50% threshold and the maximum in the local gradient methods of estimating edge location would show either an under- or overestimate of source volume depending on both the ratio of source diameter to system FWHM and the source shape. The predictions were found to be in good agreement with the measured volumes of simulated spherical sources. The use of finite-size voxels was determined to yield a discrete set of possible volumes, where the possible magnitudes were dependent on voxel size and centering of the object within the array of voxels. Decreasing voxel size was found to improve the accuracy of volume quantitation in the absence of noise.